The National Council on Compensation Insurance (NCCI) recently released two articles detailing an analysis of the The Impact of Automation on Employment – Part I, and Part II. They are researched, documented, and sourced. They lead us to the perhaps obvious conclusion that technology will change our conception and perception of work. There is room for the discussion of the "how" and the "how much," but it appears the "if" question has been answered.
In the first Part, there is a brief review of the automation advances in the early 20th Century. Obviously, those further changed America from an agrarian to manufacturing society. NCCI concludes that signs today point to a potential for a similarly significant shift resulting from "advances in artificial
intelligence and robotics." These will "potentially transform future jobs and the
structure of the labor force."
The driving force is obvious. Automation "has steadily decreased costs for decades." The decrease of cost for product inputs allows for competitive pricing of end-products. That pricing advantage renders products more appealing to consumers. The equation is neither complex nor surprising. The labor involved in manufacturing is a major component of production cost, and automation has proven an effective method of decreasing that segment of cost.
What is purportedly new in this millennium, however, is the spread of this technology leveraging into non-manufacturing environs. NCCI concludes that "technical advances in computing power, artificial
intelligence, and robotics have created the potential for automation to penetrate deeply into occupations beyond manufacturing." That technology may be further leveraged is perhaps an absolute truth, beyond argument.
However, there are those among us that remember typewriters, thesauruses, dictionaries and more. We lived through the new age of electric typewriters, correcting typewriters, memory typewriters, and the birth of electronic word processors. We lived through what we retrospectively see as slow, frustrating, and difficult programs like Word Star and Visicalc, (early) Word Perfect and SuperCalc. We recognize that anyone concluding that technology has as yet not affected "occupations beyond manufacturing" may have a short memory. By the second part, NCCI seems to acknowledge that, despite its opening hypothesis which seems to ignore it.
Knowing that tech has already touched the typing pool, secretaries, receptionists and a raft of other occupations does not change the conclusion that more change is coming. It also does not foreclose the NCCI conclusion that our future may change "more dramatically
than in the past." The NCCI conclusions are based upon studies, and identify the occupations which are likely perhaps to be the next typing pools.
One critical study, that has been mentioned in this blog before is a Ball State University (BSU) study. Of course, anyone knows that for the study of business there are a few truly elite Universities, and possibly Ball State is peerless (bias alert, the blogger earned a business degree from BSU in the Dark Ages, classes were held in a cave back then). The BSU study noted "that 87% of the job losses in manufacturing from 2000 to 2010 were due to
automation, while 13% were due to globalization and trade." There are many pundits who are quick to blame job loss on NAFTA and globalization, but it has been technology that has recently affected the greatest impact, by far.
It is noteworthy that automation's deleterious effect on jobs, has been parallel to a similarly impressive positive impact on output. The BSU study concludes that "in 2016 the United States produced almost 72% more goods
than in 1990, but with only about 70% of the workers." Inescapably, technology, automation, and robotics have impacted both the labor and consumer markets. Moore's Law suggests to us that the speed of change will increase. Previous posts on this include Salim Ismail and a Life Changing Seminar (coincidentally an NCCI seminar) and The Running Man from Pensacola.
The current predictions center on technology replacing occupations. Noted examples are "kiosks and tablets" in restaurants, "robots to process packages in warehouses," and "self-driving trucks." Each of these presents serious implications for the service industry. These are predominantly in the theme of replacing those who perform rote and routine tasks. But, the trend is now toward artificial intelligence and software that detects and then predicts patterns and performance. This intelligence will allow automation of non-routine tasks. This will displace humans even more.
NCCI also cites studies from the Kinsey Institute and the University of Oxford (neither is a Ball State, but each nonetheless has some credibility). Kinsey started with occupational data from 800 occupations and focused on 2,000 occupational functions therefrom. It then ranked these functions for susceptibility of technology replacement. The outcomes are chilling for some functions. But, there are also occupations identified that are less susceptible of automation, such as "managing
and developing people." There are functions at which humans excel and which the study does not support a strong susceptibility of computer or technology takeover. The lesson from McKinsey is that extent of technology displacement of humans will largely depend on the occupation.
That does not mean that some occupations will completely disappear while others will be totally unaffected. The suggestion is that technology will affect all occupations. That is, some portion of each will be either affected or displaced by technology. Even in occupations that are minimally susceptible to technology, there will be innovation that assists professionals therein, rendering them more efficient and effective. Thus, though those occupations will not disappear, the opportunities may be diminished in actual number or at least in growth potential.
The Oxford study was similary focused on the occupations defined by the U.S. Department of Labor. It identifies tens of occupations that are "either entirely automatable or entirely nonautomatable." The analysis from Oxford is focused upon what it calls the "three bottlenecks to
automation" which are: "perception and manipulation, creative intelligence, and social
intelligence." Once these analyses were performed on the two extremes (those "automatable or entirely nonautomatable"), the results were then projected onto all of the occupations listed by the government.
The Oxford Study provides an overall prediction that:
47% of total US employment is in occupations at high risk for automation (probabilities greater than 70%), while 19% of employment is in occupations at medium risk (probabilities between 30% and 70%), and 33% of employment is in occupations at low risk for automation (probabilities less than 30%).
NCCI concludes that each study foretells serious implications, with some occupations at greater risk than others. There is some consensus that automation change is imminent, and that it will occur sooner in the occupations with the highest risk, as the increased susceptibility to change is likely to drive both the cost incentive and the social acceptability. The impact on various localities will thus also be different because of the prevalence of those various occupations.
In the second part of this NCCI series, the analysts attempt to apply the perspectives and conclusions above to the broad spectrum of existing occupations and vocations. This is a complicated analysis because some automation will merely render current human occupations more efficient. Those changes could affect the price of such services, which likewise could create demand beyond current levels. In that paradigm, it is possible that some occupational sectors could experience growth as a result of this technological revolution. Likewise, the technology revolution could lead to occupations as yet not considered.
The alternative effect is as likely however, depending upon occupation. That is, that the expansion of technological innovation in some occupations may render human participation minimal or nonexistent. Between these two extremes, there are likely to be a spectrum of intermediate effects and outcomes. The overall effect is predicted to be marked improvement in economic output, a significant period of growth over the next decade.
NCCI predicts that overall employment will increase "about 9 million" across all segments of the economy, comparing 2014 to 2024. It is notable that compiling and analyzing data consumes time. It is apparently because of this that the baseline chosen is 2014, rather than 2017 as that latest data is likely not yet available. But note that the decade thus selected for study, 2014-2024 is almost one third over. Thus, the analyses are both likely predictive and retrospective to some degree.
NCCI predicts the most significant employment growth in "health care and social assistance." These are occupations in which technology are likely to produce efficiencies for humans, but not to replace them. This growth is also based upon the perceived increasing need for such services, likely due to an aging population and the increasingly complex nature of services. That good news is contrasted against manufacturing, with a predicted employment decline of 1.6 million. NCCI attributes this to "both to
declining employment" already perceived, and the "relatively high exposure to automation
penetration."
The prediction is that "automation penetration" might overall result in elimination of about "6.4 million without decreasing output." That will be less pervasive in the occupations that are difficult to automate, those that involve "creative and interpersonal skills or unpredictable physical tasks." These will include health care, teaching, and other social-interaction dependent functions. However, the effects will be more intense in examples like "data collection and processing as well as routine physical tasks." A chilling prediction is noted of exceptional loss of banking and financial analysis jobs in the next 5 years.
NCCI contends in the second part that the effects of technology are upon us. It concludes that it "has already
led to increasing 'job polarization.'" The impact thus far has largely affected the "the middle of the wage distribution," primarily "physical labor in manufacturing" and "white-collar
clerical jobs." These white-collar jobs being "displaced by better software and computing power," see the typing pool discussion above.
NCCI notes a perceived gender proportionality in this analysis. Data is cited regarding the perceptions of gender participation in various employment segments. While that data is intriguing, it is perhaps flawed in its reliance upon perceptions of gender. Many continue to see gender as somehow an absolute that can be defined and analyzed. There is a growing trend, or at least there is media promotion of such trend, toward a belief that gender is a mental rather than physical state. That set of progressive hypotheses and beliefs might render any supposed gender analysis unsupportable due to its assumptions being based upon physical perceptions of gender.
There is a notable set of predictions that implicate workers compensation. This is notable in both the frequency of injury and the resulting insurance rates that are charged for risk. NCCI concludes that automation could significantly impact physical occupations (in which there are risks from pushing, pulling, lifting, climbing, falling, etc.). With less humans performing such tasks, there is therefore less risk of injury from such risks. NCCI sees potential for "making work
within sectors safer" and "decreasing frequency of injuries."
There will be a variety of cost/benefit analyses performed across the spectrum of economic function, including employment. Business will contend with choices between the cost of capital investment versus labor. While automation suggests a cost savings over time, as a result of both the decreased direct (wages/benefits) and indirect (workers' compensation) labor costs, those decreases will affect the company bottom line only over time. In comparison, the investment in such automation (capital investment in facilities, equipment, software) will require more immediate, up-front, investment.
It seems likely that automation (robotics, artificial intelligence, and more) will inevitably continue to permeate and penetrate the world of goods and services around us. The questions will seemingly be about how long that penetration will require, which will seemingly be influenced by social factors such as market acceptance, financial factors such as capital cost versus existing labor paradigms, and remembering Moore's Law, the speed at which innovation improves and thus decreases in price.
An example would perhaps illuminate. McDonalds has achieved market dominance through a series of decisions. Their franchise model, standardization, and promotion have each been featured in various academic studies of success. Currently, McDonalds is installing self-serve kiosks, which will facilitate ordering. According to Forbes, these will cost each restaurant between $120,000 and $160,000. With the new year, Florida's minimum wage will be $8.25 per hour. There is at least a perception that most counter help is initially paid this rate.
A franchisee cost/benefit analysis might include the Forbes conclusions that 70% of McDonalds business is in the "drive-thru" in which kiosks are less viable (today). Thus, the kiosk investment might decrease only 30% of a store's workforce. The analysis might also include conclusions about how many kiosks are required to replace how many workers at $8.25 per hour. An investment of $160,000 would require 19,394 hours at $8.25 per hour to recoup. That is, if the investment replaced only one wage earner, then that investment would pay for itself in 19.394 hours (about 485 work weeks, over 9 years, at 40 hours each).
Of course, one kiosk might well replace more than one worker. And perhaps this $160,000 costs is for replacing multiple workers. The investment cost cited might replace two hourly employees (recouped in 242 weeks) or three (161 weeks) or four (121 weeks). Those predictions and facts (how much must be invested, and how long will be required to recoup that investment) will, in part, drive the analysis of implementing automation. And, in a manufacturing environment, a similar cost/benefit analysis might be the end of the decision process.
In the McDonalds environment, there will also be social issues for consideration. How will the customer react to the change from human interaction to automation? To some extent, we have already witnessed this. At one time, checking-in for a flight at the airport required interaction with a desk agent. About ten years ago, that evolved to using a kiosk but with desk agent assistance. That has since evolved to check-in over the Internet or with the kiosk, and little if any desk agent interaction. The overall effect has been a marked decrease in the number of desk agents.
Similarly, various restaurants have installed Ziosk tablets in the last five years. These provide entertainment (games), but also allow the customers to pay their bill. They then evolved to allow customers to interact with staff for things like re-ordering (drink refill). Then, they expanded capabilities to allow customers to place their initial food order. And now, some report, they are even able to both watch and listen to customers. Thus, we have witnessed technology arrive, and then evolve in our experiences, and in meeting our needs.
The point of these examples is that the social issue may be less persuasive than expected. While humans are persistently resistant to change (I initially and stubbornly refused to use airport kiosks and grocery store self-check outs), we are also very likely to adapt. As the self-check outs became more ubiquitous, some stores decreased the available lanes with human interaction. The lines for those check-outs were thus longer, the self-check outs shorter, and I evolved, and began to use the self-check. Currently, I gravitate to those self-check because they are quicker. With those proofs, the McDonalds' franchisee analysis of kiosks versus employees may well be more simple economics, and less "social acceptance" driven.
The NCCI analyses are well worth the read. While there are skeptics who believe that automation will not markedly impact the work force and workers' compensation, there is significant evidence to the contrary. The Impact of Automation on Employment – Part I, and Part II are worthy of consideration. The job you save might just be your own.
